Loss of Gut Barrier Integrity Triggers Activation of Islet-Reactive T Cells and Autoimmune Diabetes

Loss of gut barrier integrity triggers activation of islet-reactive T cells and autoimmune diabetes

Chiara Sorinia,b, Ilaria Cosoricha, Marta Lo Contea, Lorena De Giorgia, Federica Facciottic, Roberta Lucianòd, Martina Rocchid, Roberto Ferraresea, Francesca Sanvitod, Filippo Canduccie, and Marika Falconea,1

aExperimental Diabetes Unit, Division of Immunology, Transplantation and Infectious Diseases, IRCCS San Raffaele Scientific Institute, 20132 Milan, Italy; bImmunology and Allergy Unit, Department of Medicine, Karolinska Institute, SE-171 77 Stockholm, Sweden; cDepartment of Experimental Oncology, European Institute of Oncology IRCSS, 20139 Milan, Italy; dPathological Anatomy Unit, IRCSS San Raffaele Scientific Institute, 20132 Milan, Italy; and eDepartment of Biotechnology and Life Sciences, University of Insubria, 21100 Varese, Italy

Edited by Lawrence Steinman, Stanford University School of Medicine, Stanford, CA, and approved May 16, 2019 (received for review August 23, 2018) Low-grade intestinal inflammation and alterations of gut barrier presence of a leaky gut and the pathogenic process of T1D (21), integrity are found in patients affected by extraintestinal autoim- it is yet to be determined whether functional loss of gut barrier mune diseases such as type 1 diabetes (T1D), but a direct causal link integrity does directly trigger beta cell autoimmunity and, if it between enteropathy and triggering of autoimmunity is yet to be does, how this process occurs. It has been proposed that leakage established. Here, we found that onset of autoimmunity in preclinical of intestinal barriers could lead to uncontrolled passage into the models of T1D is associated with alterations of the mucus layer systemic circulation of bacterial components that directly mediate structure and loss of gut barrier integrity. Importantly, we showed beta cell damage and/or activate beta cell autoimmunity within that breakage of the gut barrier integrity in BDC2.5XNOD mice car- pancreatic lymph nodes (PLNs) and tissues (22). Alternatively, rying a transgenic T cell receptor (TCR) specific for a beta cell auto- autoimmune T cells specific for beta cell antigens could be acti- antigen leads to activation of islet-reactive T cells within the gut vated by bacterial products at the intestinal level and subsequently mucosa and onset of T1D. The intestinal activation of islet-reactive travel to PLNs and islets to mediate beta cell damage (23). T cells requires the presence of gut microbiota and is abolished when The gastrointestinal barrier is a fundamental gatekeeper to mice are depleted of endogenous commensal microbiota by antibiotic avoid the contact between luminal content and the human body. treatment. Our results indicate that loss of gut barrier continuity can The barrier is composed of a mucus layer and an intestinal ep- lead to activation of islet-specific T cells within the intestinal mucosa ithelial barrier (IEB), and both are crucial to prevent the passage and to autoimmune diabetes and provide a strong rationale to design of commensal bacteria, pathogens, and food antigens from the innovative therapeutic interventions in “at-risk” individuals aimed at lumen into the gut tissue and systemic circulation. The IEB is a restoring gut barrier integrity to prevent T1D occurrence. single layer of epithelial cells held together by a complex junctional system composed of tight junctions, adherent junctions, autoimmune diabetes | microbiota | gut inflammation and desmosomes. Tight junctions are composed of transmembrane proteins such as occludin, claudin, junctional adhesion molecules ype 1 diabetes (T1D) is an autoimmune disease mediated by (JAMs), tricellulin, and angulins whose interaction between Tself-reactive T cells that destroy insulin-producing beta cells themselves and with intracellular scaffolding proteins, i.e., zonula of the pancreatic islets. Although it is known that genetic and occludens proteins (ZOs) [such as tight junction protein 1 environmental factors are involved in T1D pathogenesis, the mechanisms governing the activation of islet-specific autoim- Significance mune T cells are still unclear (1). Several environmental risk factors for T1D act at the intestinal level such as enteric infec- Functional loss of gut barrier integrity with subsequent in- tions (i.e., enteroviruses and rotaviruses) (2, 3), reactions to creased antigen trafficking and occurrence of low-grade in- dietary antigens (i.e., cow’s milk and gluten) (4, 5), and testinal inflammation precede the onset of type 1 diabetes modifications of the gut microbiota induced by diet composition, (T1D) in patients and preclinical models, thus suggesting that excessive hygiene, antibiotics, and other modulators (6–9). Those these events are mechanistically linked to the autoimmune factors, specifically proinflammatory diet and alteration of the pathogenesis of the disease. However, a causal relationship microbiota composition (dysbiosis), induce intestinal in- between increased intestinal permeability and autoimmune flammation and modify the metabolic and immunological profile diabetes was never demonstrated. Our data show that break- in the intestinal mucosa of T1D patients (10, 11). In line with this age of gut barrier continuity leads to activation of islet-reactive idea, the development of clinical diabetes in patients and pre- T cells in the intestine, thus triggering autoimmune diabetes. clinical models of T1D is often preceded by clinically silent signs An important implication of our findings is that restoration of a of intestinal inflammation such as increased permeability, lym- healthy gut barrier through microbiota and diet modulation in phocyte infiltration, expression of MHC II molecules, and the diabetes-prone individuals could reduce intestinal activation of presence of inflammatory cytokines in the intestinal mucosa (12– islet-reactive T cells and prevent T1D occurrence. 18). In humans, signs of intestinal inflammation are detectable Author contributions: M.F. designed research; C.S., I.C., M.L.C., L.D.G., F.F., R.L., M.R., and before clinical onset of T1D in individuals with beta cell auto- R.F. performed research; C.S., R.F., F.S., F.C., and M.F. analyzed data; and C.S. and M.F. immunity (islet autoantibody positivity) and no hyperglycemia wrote the paper. (19). Similarly, augmented gut permeability appears before the The authors declare no conflict of interest. development of insulitis in diabetes-prone rats (BB-DP) in This article is a PNAS Direct Submission. comparison with diabetes-resistant rats (BB-DR) (12, 20). Those This open access article is distributed under Creative Commons Attribution-NonCommercial- findings indicate that the breakage of gut barrier integrity with NoDerivatives License 4.0 (CC BY-NC-ND). subsequent increased antigen trafficking and occurrence of low- See Commentary on page 14788. grade intestinal inflammation precede the onset of T1D and are 1To whom correspondence may be addressed. Email: [email protected]. directly related to its pathogenesis rather than secondary to This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. diabetes-induced metabolic alterations, i.e., hyperglycemia. However, 1073/pnas.1814558116/-/DCSupplemental. although these data suggest a causal relationship between the Published online June 10, 2019.

15140–15149 | PNAS | July 23, 2019 | vol. 116 | no. 30 www.pnas.org/cgi/doi/10.1073/pnas.1814558116 Downloaded by guest on October 1, 2021 (Tjp1)], is fundamental to maintain tight junction integrity and the integrity of the gut barrier and mucus layer architecture in control paracellular trafficking. In patients and rat models of the intestine of NOD mice. Our fluorescein isothiocyanate T1D alterations of the IEB have been reported in association (FITC)-dextran test revealed an increased intestinal permeability SEE COMMENTARY with gut inflammation (12, 24). However, the IEB is not the only in female NOD mice in comparison with different control strains intestinal barrier that is important to prevent bacterial trans- (Fig. 1A) that started at 10–12 wk before onset of autoimmune location. In fact, the first barrier that commensal bacteria en- diabetes indicating that breakage of the intestinal barrier in- counter before entering in contact with the host is the mucus tegrity is an early event in T1D pathogenesis in NOD mice as layer that covers the entire gastrointestinal canal (25). The main previously shown in T1D patients and diabetic BB rats (12, 19). function of the mucus barrier is to limit the contact between the To further assess integrity of the IEB in the intestine of NOD gut mucosa and harmful molecules present in the intestinal lu- mice, we performed RT-qPCR analysis to measure mRNA ex- men such as hydrochloric acid, digestive enzymes, and bile salts. pression relative to structural IEB proteins such as the tight However, the mucus layer does not simply act as a diffusion junction protein 1 (Tjp1) and claudin-1 (Cldn1). Our analysis barrier but plays important dynamic functions that regulate the revealed a slight but not significant decrease in the relative ex- type of commensal bacteria residing in the inner mucus layer, pression of mRNA encoding for those structural IEB proteins in allow the passage of food and bacterial products into the gut the large intestine (colon segment) of NOD compared with tissue and systemic circulation, and has important immune reg- control mice (Fig. 1B and SI Appendix, Fig. S1A). Next, we asked ulatory functions. For example, a healthy mucus architecture is whether the mucus layer structure and relative expression of fundamental to allow hosting of beneficial commensal bacteria mucin proteins are altered in the intestine of NOD mice. Immu- such as the short-chain fatty acid (SCFA)-producing bacteria nohistochemistry of the mucus layer with anti-Muc2 monoclonal (26) that promote immune tolerance in the gut and extra- antibody showed modifications of the mucus architecture and intestinal tissues (27, 28). Moreover, the mucus layer contains breakage of the mucus layer continuity in the NOD mice compared antimicrobial peptides (AMPs) and mucins, such as Muc1, with control mice (Fig. 1C and SI Appendix,Fig.S1B). Our quan- Muc2, Muc3, Muc4, Muc12, Muc13, and Muc17, that play key tification of mucin expression highlighted a significant decrease in immune-modulatory functions (29, 30). For example, Muc2 is + the relative expression of immune-regulatory mucins Muc1 and important for the cooperation between goblet cells and CD103 + Muc3andincreaseininflammatoryMuc4mucininthelargein- dendritic cells (DCs) leading to FoxP3 Treg cell differentiation testine of NOD mice compared with control mice (Fig. 1D and SI and immune tolerance in the gut (31). Muc1 and Muc3 have Appendix,Fig.S1C). We focused our analysis on the mucus archi-

strong antiinflammatory function acting as decoy molecules on INFLAMMATION

tecture in the large intestine because this is the intestinal segment IMMUNOLOGY AND the apical cell surface of enterocytes to limit bacterial adherence, where the mucus layer is more represented and plays important translocation, and intestinal inflammation (32). regulatory functions. In fact, anti-MUC2 immunohistochemistry The integrated response of these combined barrier systems, analysis showed a much thinner and less structured mucus layer i.e., the mucus layer and the IEB, is crucial to regulate the in- in the ileum (small intestine) in comparison with the colon (large teraction between commensal microbiota and immune cells intestine) segment in control mice (SI Appendix,Fig.S2). Never- within the gut mucosa and systemically (26). So far, increased gut theless, our comparative analysis revealed differences between leakiness present in animal models and T1D patients has been NOD and control mice also at the small intestinal level with a re- exclusively ascribed to alteration of the IEB, while the mucus duced thickness and breaks in the mucus layer continuity in the layer integrity and mucin composition have been poorly studied NODmiceincomparisonwithcontrolmice(SI Appendix,Fig. except for some mucin alterations found in the BB rat model of T1D (33). S2A). At the mucin level, we did not detect alterations in Muc1 and Here, we show in preclinical models of autoimmune T1D that Muc3 expression but increased Muc4 expression in the small in- the damage of the intestinal barrier involves primarily the mucus testine of NOD mice compared with control mice in line with what layer with profound modifications of its structure and composi- was observed in the large intestine (SI Appendix,Fig.S2B). The tion at the level of the small and large intestine. Most importantly, mucus layer and mucin composition play important immune- we demonstrated that induction of chronic colitis by low-dose modulatory functions that are fundamental to regulate the cross dextran-sulfate sodium (DSS), by altering gut barrier integrity talk between the gut microbiota and the host immune system. and mucus layer composition, triggers activation of islet-reactive Hence, we asked whether mucus layer modifications are associated T cells in the gut mucosa [diabetogenic T cells of T cell receptor with alteration of gut mucosal immunity in the NOD mice. In line (TCR)-transgenic (tg) BDC2.5XNOD mice] leading to autoim- with this notion, we found increased levels of several inflammatory α β mune diabetes. In our colitis-induced diabetes model, the pres- cytokines such as TNF- ,IL1-, IL-23p19, and IL-17 in the large ence of the endogenous commensal microbiota is necessary to intestinal mucosa of NOD mice compared with control mice (Fig. trigger activation of beta cell autoimmunity, and T1D occurrence 1E and SI Appendix,Fig.S1D). We studied modifications of the is abolished in DSS-colitis BDC2.5XNOD mice treated with anti- mucus structure and composition as well as the inflammatory gut biotics. Our data establish a causal relationship between breakage environment in the NOD mice early on in T1D progression (4–6wk of intestinal barriers and autoimmune pathogenesis of T1D in of age). Our analysis at a later time point (10–12 wk of age) in- preclinical models and suggest that enteropathy characterized by dicated that most changes such as increased gut permeability, increased gut permeability and modification of mucus structure thinning of the mucus layer, and alteration of Muc1 expression (but is not an epiphenomenon in T1D but directly responsible for not of Muc3 and Muc4) are maintained with age (SI Appendix,Fig. triggering beta cell autoimmunity. S3). Interestingly, we demonstrated that the increased inflammatory gut environment in the NOD mice triggered significant alter- Results ations of gut immune homeostasis in the small and large intestine of Increased Intestinal Permeability and Loss of Mucus Barrier Integrity NOD mice with augmented frequency of effector Th17 cells in both + in Diabetic NOD Mice. Alterations of the IEB have been reported districts and reduction of FoxP3 Treg cell percentages in the large in T1D patients and diabetic BB rats in association with gut in- intestine (Fig. 1F). The immune alterations of Th17/Treg cell fre- flammation (12, 24), but the gut barrier integrity has been poorly quencies in the small and large intestine of NOD mice compared studied in the NOD mice, the spontaneous murine model of with control strains were observed at 10–12 wk of age but not T1D. Moreover, the mucus layer, an important gut barrier con- earlier (4–6 wk of age), as if the proinflammatory intestinal envi- taining immunoregulatory molecules such as antimicrobial pep- ronment in the first weeks of age requires some time to significantly tides and mucins, has never been analyzed in T1D. We assessed affect gut mucosal immunity (SI Appendix,Fig.S4).

Sorini et al. PNAS | July 23, 2019 | vol. 116 | no. 30 | 15141 Downloaded by guest on October 1, 2021 Fig. 1. Increased intestinal permeability and loss of mucus barrier integrity in NOD mice. (A) FITC-dextran in vivo permeability assay in female BALB/c and NOD mice at 4, 12, and 18 wk of age. (B) RT-qPCR analysis of tight junction protein 1 (Tjp1) and claudin 1 (Cldn1) on tissue homogenates from the colons of 4- wk-old BALB/c and NOD mice. (C) Immunohistochemistry of Muc2 protein in the colons of 4-wk-old NOD mice and controls. (D) RT-qPCR analysis of Muc1, Muc2, Muc3, and Muc4 mucin genes in the colons of BALB/c and NOD mice. (E) RT-qPCR analysis of cytokine genes encoding TNF-α (Tnf), interleukin-1β (Il1b), subunit p19 of IL-23 (Il23-p19), and IL-17A (Il17a) on tissue homogenates from the colons of 4-wk-old BALB/c and NOD mice. (F) Flow-cytometric analysis of Th17 and Treg cells in the small intestinal (SI) and large intestinal (LI) lamina propria (Lp) of 12-wk-old BALB/c and NOD mice. Data are presented as mean percentages ± SD of IL-17+CD4+ and FoxP3+CD25+CD4+ T cells out of total CD4+ T cells from two independent experiments (n = 4–10 mice per group). *P < 0.05; **P < 0.01; ***P < 0.001.

Chronic Colitis and Breakage of Intestinal Barrier Integrity Trigger (BDC2.5 T cells) are activated, for example through islet-antigen Activation of Islet-Reactive T Cells in the Gut Mucosa. Our data in- presentation in the target tissue (37, 38). In those mice, we al- dicate that increased permeability and alterations of immune tered the gut barrier continuity and, specifically, the mucus ar- homeostasis are present in the intestinal mucosa of NOD mice chitecture by administration of low-dose DSS, according to a and are associated with important changes in the mucus layer protocol that specifically targets the mucus layer integrity (39) architecture and composition. Next, we wanted to establish a (SI Appendix, Fig. S5A), and then analyzed the activation state causal link between the presence of a “leaky” gut, breakage of and cytokine secretion profile of islet-reactive BDC2.5 T cells + + the mucus layer structure, and activation of beta cell autoim- (Vβ4 CD4 T cells). BDC2.5XNOD mice undergoing DSS munity. The observation that lymphocytes isolated from mes- treatment developed chronic colitis, characterized by significant enteric lymph nodes (MLNs) of young prediabetic NOD mice weight loss and colon shortening, as well as increased gut per- are capable to transfer autoimmune diabetes suggests that the meability and alterations of tight junction protein expression (SI intestine could be a preferential site for activation and expansion Appendix, Fig. S5 B–F). Chronic colitis in those mice correlated of islet-reactive T cells (34). Hence, we hypothesized that en- with substantial modifications of the mucus structure (Fig. 2A) teropathy and breakage of the intestinal barrier integrity are not and mucin protein composition (Fig. 2B). Specifically, we de- epiphenomena but could play a pathogenic role in T1D patho- tected a mucin RNA expression profile comparable to that ob- genesis through triggering intestinal activation of islet-reactive served in the intestine of NOD mice with a significant decrease T cells. In the NOD mouse model, due to the low number of in mRNA expression of immune-regulatory mucins Muc1 and circulating islet-reactive T cells, it is not possible to track them Muc3 (Fig. 2B). Next, to assess the effect of loss of gut barrier and assess their activation state in peripheral lymphoid organs integrity on islet-reactive T cell activation, we analyzed the ex- + and tissues. To test our hypothesis and monitor the activation pression levels of the CD44 activation marker on Vβ4 BDC2.5 state of islet-reactive T cells, we used a TCR-tg murine model T cells in the intestinal mucosa of DSS-colitis BDC2.5XNOD + (BDC2.5XNOD mice) in which majority of CD4 T cells (∼90%) mice (see SI Appendix, Fig. S6, for gating strategy). Our data express a tg TCR specific for a beta cell antigen (35). Tg show that chronic DSS-colitis and breakage of the intestinal + BDC2.5XNOD mice bear rearranged genes for the TCR-α (Vα1) barriers increased the percentage of activated CD44 BDC2.5 + + and TCR-β (Vβ4) chains from an islet-reactive CD4 T cell clone (Vβ4 ) T cells within the gut mucosa (Fig. 2C). Moreover, we (BDC2.5 T cell clone) isolated from a diabetic NOD mouse (36). found that the cytokine-secretion phenotype of islet-reactive Those tg mice, despite carrying a large diabetogenic T cell rep- BDC2.5 T cells in the intestine was skewed toward an effector ertoire and having large T cell infiltrates in the pancreatic islets, Th1/Th17 phenotype similarly to that observed in the gut of do not develop autoimmune diabetes unless islet-reactive T cells NOD mice (Fig. 2D). This bias was associated and possibly related

15142 | www.pnas.org/cgi/doi/10.1073/pnas.1814558116 Sorini et al. Downloaded by guest on October 1, 2021 SEE COMMENTARY

Fig. 2. Chronic colitis and breakage of intestinal barrier integrity trigger activation of islet-reactive T cells in the gut. (A) Low-dose DSS administration induced alteration of mucus structure as shown by immunohistochemistry analysis with anti-Muc2 mAb in the colon of control and DSS-colitis BDC2.5XNOD mice (DSS). (B) RT-qPCR analysis revealed alteration of mucin genes’ expression in the colon of DSS-colitis BDC2.5XNOD mice with reduced expression of Muc1, + + − + Muc2, and Muc3 mRNA. (C) Percentages of activated CD44hi T cells among tg Vβ4 CD4 BDC2.5 T cells (Left) and non-tg Vβ4 CD4 T cells (Right) in intestinal lymphocytes isolated from colon of control (ctrl) and DSS-colitis (DSS) BDC2.5XNOD mice. (D) Representative flow cytometry plots (Left) and bar graph with + + mean percentages ± SD (Right) of cytokine-producing CD4 T cells (out of CD45 cells) within the gut mucosa of control (ctrl) and DSS-colitis BDC2.5XNOD mice (two independent experiments; n = 8 mice per group). (E) Cytokine bead array (CBA) assay for IL-17A and IFN-γ production by lymphocytes from the intestine of control (ctrl) and DSS BDC2.5XNOD mice stimulated in vitro with Con A for 72 h. Data are presented as mean percentages ± SEM of IL-17+CD4+ + + + + INFLAMMATION and FoxP3 CD25 CD4 T cells out of total CD4 T cells from triplicate wells. Data are from one representative experiment out of two. *P < 0.05; **P < 0.01; IMMUNOLOGY AND ***P < 0.001.

to the presence of a highly inflammatory cytokine milieu enriched trigger beta cell damage and autoimmune diabetes unless they for IL-23 (a Th17-inducing cytokine) in the gut mucosa of DSS- are activated by viruses, cytokine milieu, etc. (37, 38). In accor- colitis BDC2.5XNOD mice (SI Appendix,Fig.S5E). The activation dance with previous findings, in our control BDC2.5XNOD mice, state and strong effector phenotype (Th1/Th17 cells) of BDC2.5 we detected large infiltrates around healthy islets (periinsulitis) T cells was confirmed by their increased secretion of IL-17 and with no signs of insulitis (Fig. 3E, Left). On the contrary, the IFN-γ upon in vitro stimulation (Fig. 2E). These data demonstrate pancreatic islets of DSS-colitis BDC2.5XNOD mice were com- that breakage of intestinal barrier integrity and alteration of mu- pletely invaded by lymphocytes with very limited residual beta cus structure in DSS-colitis BDC2.5XNOD mice are associated cell mass (Fig. 3E, Right). Importantly, in line with the histo- with activation of islet-reactive BDC2.5 T cells within the gut logical signs of islet damage, we observed that 60% of DSS- mucosa. colitis BDC2.5XNOD mice developed autoimmune diabetes, while 100% of their healthy littermates remained diabetes-free Islet-Reactive T Cells Activated in the Gut Mucosa Migrate to PLNs (Fig. 3F). These data indicate that increased activation and ac- and Tissues and Induce Autoimmune Diabetes. Previous reports in- quisition of an effector Th1/Th17 phenotype by islet-reactive dicated that an immunological axis between the gastrointestinal BDC2.5 T cells in the intestine of DSS-colitis BDC2.5XNOD tract and pancreatic tissues exists (23), so that T cells continu- mice enhance their aggressiveness, thus enabling them to trigger ously travel from the intestinal mucosa to pancreatic tissues and beta cell destruction and autoimmune diabetes. into the systemic circulation. Hence, we asked whether islet- reactive BDC2.5 T cells, after being activated in the intestinal Induction of Autoimmune Diabetes Requires the Cross Talk Between mucosa, travel to PLN and islets to induce beta cell damage and Islet-Reactive T Cells and the Gut Commensal Microbiota. Next, we autoimmune diabetes. As expected, we found that most of the asked which are the mechanisms that trigger activation of islet- islet-reactive BDC2.5 T cells present in the PLN and intraislet reactive BDC2.5 T cells in the gut mucosa of DSS-colitis lymphocytes (IILs) of BDC2.5 mice express the gut homing BDC2.5XNOD mice. First, we tested whether DSS-colitis af- marker α4β7 integrin, a proof of their intestinal origin, even in fected the composition of gut microbiota favoring overgrowth of the absence of DSS-colitis (Fig. 3A). This finding is not sur- proinflammatory bacterial species that could be responsible for prising since it is known that T cells continuously travel from the activation and acquisition of an effector phenotype by islet- intestinal mucosa to pancreatic tissues through the gut–pancreatic reactive BDC2.5 T cells in DSS BDC2.5XNOD mice. By ultra- axis (23). However, it is important to note that BDC2.5XNOD mice deep pyrosequencing of barcoded 16S rRNA gene amplicons, we + with DSS-colitis showed a higher absolute number of α4β7 islet- analyzed the gut microbiota community in stool samples of DSS- reactive BDC2.5 T cells in the PLN in comparison with their colitis and healthy BDC2.5XNOD mice at 4 wk after the first healthy littermates (Fig. 3B). Moreover, BDC2.5 T cells found in DSS administration when the animals showed early signs of co- PLN and islet infiltrates of DSS-colitis BDC2.5XNOD mice showed litis (SI Appendix, Fig. S5B) but were still normoglycemic (Fig. an activated CD44highCD62low phenotype (Fig. 3C)andabiased 3F). Although we normalized bacterial load between samples effector Th17 cytokine-secretion phenotype compared with their based on total DNA and we cannot infer on differences in ab- counterparts from healthy mice (Fig. 3D). solute numbers of live commensal bacteria, we found substantial In BDC2.5XNOD mice, large infiltrates of islet-reactive differences in the alpha-diversity (Fig. 4A) and relative abun- BDC2.5 T cells surround the pancreatic islets, but they do not dance of some bacterial strains in the DSS-colitis mice compared

Sorini et al. PNAS | July 23, 2019 | vol. 116 | no. 30 | 15143 Downloaded by guest on October 1, 2021 to the gut mucosa and gut-associated lymphoid tissue (GALT) and favor the cross talk between the commensal microbiota and the immune system, thus leading to activation of islet-reactive T cells. In support of bacterial translocation as a consequence of damage of gut barrier integrity, we found commensal bacteria in the MLNs of DSS-treated BDC2.5XNOD mice (SI Appendix, Fig. S7A). To test whether the gut commensal bacteria is required to induce autoimmune diabetes in DSS BDC2.5XNOD mice, we depleted those mice of endogenous commensal microbiota before induction of chronic DSS-colitis. To this aim, we treated BDC2.5XNOD mice with a broad-spectrum antibiotic mixture (ampicillin, metronidazole, neomycin, and vancomycin) that induces complete depletion of endogenous microbiota (SI Ap- pendix, Fig. S7B). Importantly, previous reports showed that the antibiotic treatment depletes the gut microbiota without affecting the DSS-induced damage of the gut barrier integrity in DSS- treated mice (40). Remarkably, none of the DSS BDC2.5XNOD mice depleted of their commensal gut microbiota developed T1D. These results demonstrate that loss of gut barrier integrity and intestinal inflammation are not sufficient to induce T1D and the presence of commensal gut microbiota is required to trigger intestinal activation of islet-reactive BDC2.5 T cells (Fig. 5B). Our results indicate that the gut microbiota is necessary but not sufficient to induce T1D, but how does the commensal bacteria trigger beta cell autoimmunity upon breakage of the intestinal barriers? Loss of intestinal barrier integrity could lead Fig. 3. Islet-reactive BDC2.5 T cells activated in the gut elicit pancreatic to passage of bacterial products that could stimulate islet- autoimmunity. (A) FACS analysis of expression levels of α4β7 integrin on reactive T cells through a TCR-mediated mechanism. In fact, + CD4 T cells in the pancreatic lymph nodes (PLNs) and intraislet lymphocyte commensal bacterial strains may contain molecules that mimic + (IIL) compartment of BDC2.5XNOD mice. (B)Absolutenumbersofα4β7 TCR-tg + + − + self-antigens and directly trigger TCR-mediated activation of islet-reactive BDC2.5 T cells (Vβ4 CD4 T cells) and non-tg T cells (Vβ4 CD4 islet-reactive T cells as recently demonstrated for a diabetogenic + T cells) in the PLNs of DSS-colitis (DSS) and control (ctrl) BDC2.5XNOD CD8 T cell clone (41). Alternatively, bacterial products could mice. (C) Representative flow cytometry plots (Left) and percentage (Right) hi − bind pattern recognition receptors on intestinal DCs and in- of activated CD44 CD62L T cells in the PLN and IIL of control and DSS BDC2.5XNOD mice. (D) Absolute numbers of Th17 cells in the PLN of control crease their capacity to stimulate all T cells in an antigen- and DSS BDC2.5XNOD mice. (E) Hematoxylin and eosin staining of pancreatic independent manner (42). To test whether the commensal gut tissue showing large lymphocyte infiltrates surrounding the pancreatic islets microbiota can trigger activation of islet-reactive T cells, we of control BDC2.5XNOD mice (Left) and massive islet infiltration in the performed in vitro experiments in which total splenocytes from DSS BDC2.5XNOD mice (Right). (F) Incidence of autoimmune diabetes in BDC2.5XNOD mice or wild-type (wt) mice were challenged with control and DSS-colitis (DSS) BDC2.5XNOD mice (n = 7 mice per group). *P < 0.05; bacterial suspensions isolated from the intestine of DSS-colitis or + **P < 0.01. healthy mice and the acquisition of an activated IFN-γ pheno- + + type on islet-reactive BDC2.5 T cells (Vβ4 CD4 ) (from tg + BDC2.5XNOD mice) or polyclonal CD4 T cells (from wt mice) with their healthy littermates. Specifically, we found that, at the + was evaluated. Our data show that CD4 BDC2.5 T cells are phylum level, Firmicutes and Deferribacteres were increased, + activated and acquired an activated IFN-γ phenotype when whereas Bacteroidetes strains were significantly less represented stimulated with commensal microbiota strains from healthy in DSS-colitis BDC2.5XNOD mice compared with their healthy BDC2.5XNOD mice (Fig. 5C). No stimulation was detected on + littermates (Fig. 4B). The family-level analysis revealed that the polyclonal CD4 T cells (splenocytes of wt mice) challenged in Bacteroidetes strains reduced in DSS-colitis BDC2.5XNOD mice vitro with the same bacterial suspensions (Fig. 5C). We cannot were Prevotellaceae, Rikenellaceae, and S24-7, while others such exclude an indirect effect of the bacterial lysates on other im- as Porphyromonadaceae were increased (Fig. 4C). Deferribactera mune cells contained in total splenocytes (e.g., DCs and mac- ceae and Erysipelotrichaceae were also augmented in DSS-colitis rophages). However, the observation that commensal bacteria BDC2.5XNOD mice (Fig. 4C). triggered IFN-γ secretion on BDC2.5 T cells from TCR-tg mice + To test whether modifications of the gut microbiota induced and not polyclonal CD4 T cells from wt mice suggests that the by DSS-colitis were directly responsible for intestinal activation commensal bacterial strains contained in the healthy murine of islet-reactive T cells, we performed fecal transfer experiments intestine are capable to activate autoimmune islet-reactive by orally gavaging stool suspensions from DSS BDC2.5XNOD BDC2.5 T cells with an antigen-specific TCR-mediated mecha- mice into recipient BDC2.5XNOD mice previously depleted of nism. In support to the TCR dependency of the microbiota- their endogenous gut microbiota through antibiotic treatment. induced stimulation, we found that addition of anti-MHC class Our data show that oral transfer of gut microbes from DSS- II mAb together with fecal material to the BDC2.5 T cell cul- + colitis diabetic donors into healthy BDC2.5XNOD recipients tures reduced their acquisition of an IFN-γ phenotype (SI Ap- was not sufficient to activate beta cell autoimmunity and induce pendix, Fig. S8). These data suggest that interaction between T1D (Fig. 5A). Although the fecal transfer does not lead to components of the endogenous commensal bacteria and T cells complete colonization of the host with all donor bacterial spe- that occurs in the intestinal mucosa upon breakage of the gut cies, our data suggest that modifications of the gut microbiota barrier integrity could potentially lead to activation of islet- induced by DSS-colitis were not sufficient to induce beta cell reactive T cells through a TCR-mediated mechanism. We also autoimmunity. Next, we searched the mechanism through which observed that bacterial lysates from commensal gut microbiota loss of gut barrier integrity could promote autoimmune diabetes. collected from DSS BDC2.5XNOD mice were more stimulatory + Damage of the gut barrier could lead to bacterial translocation on CD4 BDC2.5 T cells (Fig. 5C) compared with bacterial

15144 | www.pnas.org/cgi/doi/10.1073/pnas.1814558116 Sorini et al. Downloaded by guest on October 1, 2021 SEE COMMENTARY INFLAMMATION Fig. 4. DSS-colitis alters the composition of the commensal microbiota in BDC2.5XNOD mice. (A) Alpha diversity of gut microbiota from control (ctrl) and IMMUNOLOGY AND DSS-colitis (DSS) BDC2.5XNOD mice as analyzed with QIIME software using a phylogenic diversity (PD) tree. (B) Phylum and class-level phylogenetic classi- fication of 16S rRNA cDNA frequencies in the mucosa and luminal content of the intestine of control and DSS-treated BDC2.5/NOD mice. (C) 16S sequencing comparisons of DSS-colitis vs. control BDC2.5XNOD mice. Bacterial families and genera shown represent those found to be significantly different in the pairwise comparison. *P < 0.05; **P < 0.01.

strains from healthy BDC2.5XNOD mice. Although the differ- barrier that commensal bacteria encounter before entering in ence was not significant (P = 0.05), the latter observation sug- contact with the host and plays important dynamic functions that gests that dysbiosis induced by intestinal inflammation could lead regulate the type of commensal bacteria that reside in the inner to selection of bacterial species that are more stimulatory for mucus layer, the passage of food and bacterial components from islet-reactive BDC2.5 T cells. the gut lumen into the mucosal tissue, and the interaction between commensal microbiota and immune cells. For example, Discussion commensal bacteria that play important immune regulatory Recent evidence suggests that the intestinal environment and, functions such as the SCFA producers reside within the mucus specifically, modifications of the microbiome profile, regulate layer (26). In addition, the mucus layer contains mucins and the pathogenesis of T1D by inducing intestinal inflammation and antimicrobial peptides (AMPs) that are released by intestinal increasing gut permeability (43, 44). Although it has been pro- cells (enterocytes, goblet cells, and Paneth cells) and play im- posed that T1D originates in the intestine (45), it remains to be portant immune-regulatory functions (30). Expression of AMPs determined how gut leakiness in T1D patients and animal such as α-defensins and cathelicidins and the regenerating islet- models promotes beta cell autoimmunity. Here, we demonstrate derived protein (Reg)IIIγ are altered in T1D. Previous studies that loss of gut barrier integrity increases activation of islet- have shown that cathelicidins are reduced in the serum of T1D reactive T cells and triggers autoimmune diabetes. patients (47) and in NOD mice dysbiosis is associated with re- The gut barrier that prevents translocation of bacterial com- duced local expression of cathelicidin-related antimicrobial ponents in the intestinal tissue is composed of a mucus layer and peptide (CRAMP) by pancreatic endocrine cells (22). Our re- an intestinal epithelial layer (IEB). The IEB regulates intestinal sults demonstrate that also the expression of mucins is altered in permeability through a well-organized system of intercellular NOD mice. In fact, we found that expression of immune- connections between the cells, composed of tight junctions, ad- regulatory Muc1 and Muc3 is reduced while expression of herent junctions, and desmosomes. The IEB is fundamental to proinflammatory Muc4 is increased in the intestine of NOD regulate the trafficking of bacterial products and metabolites and mice. Those alterations were detected at 4 wk of age, an early is damaged in patients and rat models of T1D (12). In pre- time point in T1D progression that is concomitant with the diabetic NOD mice, we found that increased gut permeability is generation of the autoimmune T cell repertoire (48). Later on associated with damage of the mucus layer and composition and (10–12 wk of age), changes in mucin expression were partially not of the IEB. Although mucus structure and composition were corrected with decreased Muc1 expression but normalized levels never studied in NOD mice, it was recently reported that, in of Muc3 and Muc4, an effect possibly due to compensatory those mice, gut colonization with Akkermansia muciniphila mechanisms. Dysbiosis and reduction of immune-regulatory counterregulates autoimmune diabetes by increasing mucus mucins associated with damage of the mucus layer could ulti- production and secretion of the antibacterial peptide RegIIIγ in mately lead to alterations of gut immune homeostasis with in- the mucus layer (46), thus suggesting that the mucus barrier plays creased Teff/Treg cell ratio and chronic inflammation that we an important role in T1D prevention. The mucus layer is the first observed in prediabetic NOD mice. AMPs maintain gut barrier

Sorini et al. PNAS | July 23, 2019 | vol. 116 | no. 30 | 15145 Downloaded by guest on October 1, 2021 those intestinal alterations and induction autoimmune diabetes was never established. Our data demonstrate that loss of gut barrier integrity and modifications of the mucus structure and composition, associated with chronic colitis (by low-dose DSS administration), trigger T1D. A previous report showed that disruption of the gut barrier induced by Citrobacterium rodentium colonization in NOD mice increased the percentage of islet- + reactive CD8 T cells in the PLN and accelerated onset of insulitis (52). Our data demonstrate that loss of gut barrier integrity is not only an accelerator in T1D pathogenesis but a causal factor so that chronic DSS-colitis triggers T1D in TCR-tg BDC2.5XNOD mice that otherwise do not develop diabetes. BDC2.5XNOD mice, despite having 90% of the circulating T cells specific for an islet antigen, do not become diabetic and can be considered like genetically susceptible individuals carrying an enlarged islet-reactive T cell repertoire and not developing T1D unless a triggering event activates beta cell autoimmunity. In BDC2.5XNOD mice, different triggering events have been reported such as viral infections (e.g., Cox- sackie infection) or any factor that perturbs the islet environment leading to inflammation, tissue damage, and the release of sequestered islet antigen resulting in the stimulation of resting islet-reactive T cells (38). Here, we show that breakage of the gut barrier is one of those triggering events that unleashes beta cell autoimmunity and provokes islet destruction in autoimmune-prone individuals. What is the mechanism that triggers activation of islet-reactive T cells in our DSS-colitis T1D model? One possibility is that bacterial components could have translocated through the intestinal mucosa into the bloodstream and pancreatic islets and mediate tissue damage, exposure of self- antigens, and local activation of islet-reactive T cells (22). Al- ternatively, islet-reactive T cells could have been activated in the gut mucosa and travel to the PLN and islets to mediate tissue damage and T1D (23). Our data support the latter hypothesis by showing that DSS-colitis increased the percentage of activated islet-reactive T cells already in the intestinal tissue of + BDC2.5XNOD mice and large part of activated Vβ4 T cells found in the PLN and within the islets are of intestinal origin and express the α4β7 marker. Our data are in accordance with a Fig. 5. Commensal gut microbiota directly triggers activation of islet- recent report showing that myelin-reactive T cells are activated reactive T cells and is necessary for diabetes induction in DSS-colitis in the GALT, and from there they travel to peripheral tissue, i.e., BDC2.5XNOD mice. (A) Diabetes-free survival of control (ctrl), DSS-colitis the central nervous system, to mediate extraintestinal autoim- (DSS) BDC2.5XNOD mice, and healthy BDC2.5XNOD mice receiving fecal mune diseases (53). Which are the mechanisms triggering acti- microbiota transplants from DSS mice (DSS-FMT). (B) Diabetes-free survival of DSS-colitis BDC2.5XNOD mice treated (DSS+Abx) or not (DSS) with an vation of islet-reactive T cells in the gut of DSS-colitis antibiotic mixture (0.5 g/L vancomycin, 1 g/L ampicillin, 1 g/L metronidazole, BDC2.5XNOD mice? Gut inflammation could lead to pre- and 1 g/L neomycin) to deplete endogenous commensal microbiota. (C) sentation of sequestered self-antigens that are already present in Fecal material from DSS-colitis (DSS-FM) or healthy (crtl-FM) BDC2.5XNOD the intestinal tissue. For example, chromogranin A, an islet- mice was added to splenocytes of control non-tg mice (ctrl T cells) or TCR-tg antigen that is capable to stimulate the diabetogenic BDC2.5 BDC2.5XNOD mice (BDC2.5 T cells), and after 96 h T cell activation was + T cell clone (54), is expressed in the gastro-entero-pancreatic assessed by measuring percentage of IFN-γ cells. Data are presented as system (55, 56) and, under chronic inflammatory conditions, mean percentages ± SEM of IFN-γ+CD4+ T cells out of total CD4+ T cells. One < < < could be presented to islet-reactive T cells in the intestine. Our representative experiment out of two is shown. *P 0.05; **P 0.01; ***P data do not support this idea by showing that gut inflammation 0.001. by itself in DSS-BDC2.5XNOD mice depleted of endogenous microbiota is not capable to activate beta cell autoimmunity. integrity by regulating expression of mucins and TJ proteins Although in those mice the antibiotic-induced bacterial de- through the WNT pathway (49). For example, REG3γ-deficient pletion may have not lasted through the entire duration of the mice harbor an altered mucus layer and increased mucosal in- experiment, the observation that microbiota-depleted mice did flammatory response to the commensal gut microbiota (50). not develop T1D suggests that the commensal gut microbiota are required for the intestinal activation of diabetogenic T cells. The Similarly, CRAMP deficiency in mice leads to a more severe gut microbiota has a strong impact in T1D pathogenesis as form of DSS-induced colitis (51). Hence, it is possible to spec- demonstrated both in humans and preclinical models (57, 58), ulate that in the NOD mice the defective release of AMPs but it is still unclear how commensal bacteria modulate beta cell previously reported (22) is responsible for the damage of gut autoimmunity. In fact, while in other autoimmune diseases such barrier integrity and the alterations of mucus structure and as rheumatoid arthritis in HLA-B27 tg rats and experimental composition that we observed. autoimmune encephalomyelitis, the preclinical models of mul- The presence of chronic gut inflammation and loss of gut tiple sclerosis, the gut microbiota plays a clear triggering role (53, barrier integrity in patients and preclinical models of T1D has 59), in the NOD mice, the spontaneous model of T1D, this role been known for long time but a causal relationship between is still controversial. In some cases, it plays a beneficial effect, for

15146 | www.pnas.org/cgi/doi/10.1073/pnas.1814558116 Sorini et al. Downloaded by guest on October 1, 2021 example in germ-free NOD mice in which the absence of gut showed that bacterial products stimulate intestinal DC within the commensal bacteria enhances T1D incidence (6). On the con- gut mucosa through pattern recognition receptors and induce an trary, in NOD mice housed in spontaneous monoculture with a innate milieu with an adjuvant-like effect that activates islet- SEE COMMENTARY Gram-positive aerobic spore-forming rod, T1D has a delayed reactive T cells and trigger diabetes (42). Although our results onset and reduced incidence (60), thus suggesting that com- are not conclusive, they support the first hypothesis by showing mensal strains are important to trigger beta cell autoimmunity. A that the islet-reactive BDC2.5 T cells are directly stimulated by possible explanation for those controversial findings is that the commensal gut microbiota from the intestine of healthy mice NOD mice represent a complex and multifactorial animal model through a TCR-mediated mechanism. We also observed that the of autoimmune diabetes. In fact, in NOD mice the generation of commensal microbiota from the inflamed intestine of DSS- the autoimmune islet-reactive T cell repertoire is strongly af- treated BDC2.5XNOD mice was slightly higher stimulatory; fected by exposure to any bacterial antigens early in life. In those however, additional experiments are necessary to assess whether – mice, the microbial environment in specific-pathogen free DSS-colitis promoted overgrowth of proinflammatory bacterial housing facility but also exposure to microbial stimuli, such as species that were more effective in stimulating islet-reactive injection with mycobacteria (61, 62) or various microbial anti- T cells in vitro in comparison with microbial suspensions from gens (63, 64), prevent beta cell autoimmunity by affecting the healthy BDC2.5XNOD mice. overall maturation of the immune system and/or with a mecha- Our study provides direct evidence that increased gut per- nism of immune deviation. The TCR-tg BDC2.5 murine model meability plays an important pathogenic role in T1D possibly differs from NOD mice because it already has a large islet- by favoring the interaction between islet-reactive T cells and reactive autoimmune T cell repertoire (35). Hence, in those commensal microbes. The murine model that we used, the mice, the confounding effect of the microbial antigens on the BDC2.5XNOD mice, have an enlarged islet-reactive T cell rep- generation of the autoimmune T cell repertoire is not present, ertoire and can be compared with “at-risk” humans with high and we could assess the role of the commensal gut microbiota in genetic susceptibility to T1D and beta cell autoimmunity (islet activation/expansion of islet-reactive T cells. Importantly, we autoantibodies positivity) in which T1D does not occur unless showed not only that, in DSS-treated BDC2.5XNOD mice, upon the diabetogenic islet-reactive T cells are activated in the pe- damage of the gut barrier, the commensal microbiota was re- riphery by a triggering event. In diabetes-prone humans, as in quired to trigger activation of islet-reactive T cells, but also BDC2.5XNOD demonstrated that microbial antigens contained in bacterial mice, the loss of gut barrier integrity could ultimately lead to intestinal activation of islet-reactive T cells, beta suspensions isolated from the healthy murine intestine are ca- INFLAMMATION pable to directly stimulate the islet-reactive T cell clone in vitro cell destruction, and T1D. An important implication of these IMMUNOLOGY AND with a TCR-mediated mechanism. Our observation is aligned findings is that restoration of a healthy gut barrier through with a previous report showing that autoreactive T cells specific microbiota and diet modulation in diabetes-prone individuals for a retinal antigen are directly activated by the commensal gut could ultimately reduce intestinal activation of islet-reactive microbiota through their clonotypic TCR and with a mechanism T cells and prevent T1D occurrence. that is not dependent on an endogenous source of the cognate Materials and Methods antigen (65). Recent findings demonstrated that translocation of commensal microbes into intestinal tissue is a common event and Mice. Female NOD mice were purchased from Charles River Laboratories. TCR- gut pathobionts are found in the GALT of healthy, un- tg BDC2.5XNOD mice were a gift from Dr. Jonathan Katz (Cincinnati Children’s Research Foundation, University of Cincinnati, Cincinnati, OH). All manipulated mice (66). Hence, commensal gut microbes can mice were maintained under specific-pathogen–free conditions in the ani- activate autoimmune T cells specific for different extraintestinal mal facility at San Raffaele Scientific Institute, and all experiments were self-antigens that circulate through intestinal tissue and GALT conducted in accordance with the Institutional Animal Care and Use Com- even in the absence of gut inflammation and damage of the gut mittee in accordance with the rules of the Italian Ministry of Health. barrier. This notion is supported by our finding that islet-reactive T cells with an activated/memory phenotype are present in the Induction of Chronic DSS Colitis. The 2% (wt/vol) DSS (molecular weight: intestine of healthy BDC2.5XNOD mice. We speculate that, 36,000–50,000 Da; MP Biomedicals) was administered to BDC2.5XNOD mice under inflammatory conditions, in our DSS-colitis model, this according with the chronic colitis protocol described by Wirtz et al. (68). mechanism was enhanced by damage of the gut barrier resulting Briefly, 4-wk-old BDC2.5XNOD mice received the DSS orally in drinking in an increased percentage of self(islet)-reactive T cells that are water for three cycles of 7 d with intervals of 2 wk. In some experiments, activated within the intestinal tissue and migrate to pancreatic BDC2.5XNOD mice were pretreated with antibiotics (0.5 g/L vancomycin, 1 g/L islets to provoke T1D. Additionally, the inflammatory gut envi- ampicillin, 1 g/L metronidazole, and 1 g/L neomycin; all from Sigma- Aldrich) for 2 wk before each cycle of DSS administration and during the ronment characterized by the presence of proinflammatory cy- DSS treatment to deplete endogenous commensal microbiota according to a tokines such as IL-23 drove islet-reactive T cells toward an well-established protocol (69). Bacterial depletion was verified by qPCR with effector Th17 cell phenotype, thus increasing their aggressive- universal eubacteria primers as previously reported (70). Mice were either ness and capacity to destroy pancreatic beta cells similarly to monitored for diabetes occurrence or killed at 8–10 wk of age for immu- what was reported in the autoimmune uveitis model (65). nological and RT-qPCR analyses. Body weight was monitored daily during The commensal gut microbiota could activate self(islet)- induction phases and every other day during clinical remission. reactive T cells through different mechanisms. Low-affinity autoreactive T cell clones are highly promiscuous, and their In Vivo Permeability Assay. Intestinal permeability was determined by FITC- TCRs can be stimulated by microbiome-derived peptides (67) dextran assay as previously described (52). Briefly, 20 mL/kg of body including agonist peptides derived from gut commensal micro- weight of PBS containing 25 mg/mL FITC-conjugated dextran (FITC-dextran; biota (65). This hypothesis is supported by the recent finding that molecular mass, 4.4 kDa; FD4, Sigma-Aldrich) was administered to each three microbial peptides derived from commensal bacteria were mouse by oral gavage. After 4 h, blood was collected and added immedi- found to be able to stimulate the diabetogenic IGRP-specific ately to a final concentration of 3% citrate-phosphate-dextrose solution + (Sigma-Aldrich). The blood samples were centrifuged (10,000 rcf at 4 °C) for NY8.3 CD8 T cell clone and intestinal colonization with 10 min, and plasma was collected. Fifty microliters of plasma were mixed Fusobacteria containing those mimicking peptides accelerated with an equal volume of PBS (pH 7.4) and added to a 96-well microplate. The diabetes incidence, thus suggesting that commensal microbes concentration of fluorescein was determined by spectrophotofluorometry could promote T1D through molecular mimicry (41). Alterna- (Wallac Victor; Perkin-Elmer Life Sciences) with an excitation wavelength of tively, the commensal gut microbiota could activate islet-reactive 485 nm and an emission wavelength of 530 nm using serially diluted samples T cells through TCR-independent mechanisms. A recent report of the FITC-dextran marker as standard.

Sorini et al. PNAS | July 23, 2019 | vol. 116 | no. 30 | 15147 Downloaded by guest on October 1, 2021 Cell Isolation and in Vitro Stimulation Experiments. Intestinal mononuclear Histology and Immunohistochemistry. For analysis of the mucus layer structure cells were isolated from the small and large intestinal lamina propria as segments of the distal colon were fixed in methanol-Carnoy’s fixative (60% previously described (71, 72). Briefly, after the removal of Peyer’s patches, dry methanol, 30% chloroform, and 10% acetic acid), washed in absolute small and large intestines were flushed with PBS, opened longitudinally, and methanol, ethanol, and xylene, embedded in paraffin wax, and sectioned at incubated twice with 5 mM EDTA and 1 mM DTT for 20 min at 37 °C to 6 μm. The sections were stained with primary anti-Mucin-2 antibody (clone remove epithelial cells and adipose tissue. Then, the intestines were cut into Sc-15334; Santa Cruz Biotechnology) and secondary donkey anti-rabbit small pieces and digested in HBSS containing 0.5 mg/mL Collagenase D IgG antibody. (Roche Diagnostics), 1 mg/mL Dispase II (Roche Diagnostics), and 5 U/mL To assess histopathological signs of diabetes, pancreata were fixed in 10% DNase I (Sigma-Aldrich) for 20 min at 37 °C in a shaking incubator. The formalin, embedded in paraffin, and sectioned into 4-μm slices. The pancreas digested tissues were washed, resuspended in 5 mL of 40% Percoll (Sigma- sections were then stained with hematoxylin–eosin and scored for insulitis. Aldrich), and overlaid on 2.5 mL of 80% Percoll in a 15-mL Falcon tube. Percoll gradient separation was performed by centrifugation at 1,000 × g for Real-Time qPCR. Immediately after the mice were killed, colons were flushed 20 min at 20 °C. The interface cells were collected and used as intestinal μ lymphocytes for fluorescence-activated cell sorting (FACS) analysis. Splenocytes with PBS, opened longitudinally, and placed in 500 L of TRIzol reagent (Life and lymph node cells were isolated by mechanical disruption of the tissues. Technologies). After homogenization with TissueRuptor (QIAGEN), RNA was μ Pancreatic IILs were isolated by a three-step digestion with 1 mg/mL Collage- extracted by adding 100 L of chloroform, precipitating the aqueous phase μ nase IV (Gibco) followed by extensive washing with HBSS containing 5% FBS. with 300 L of 70% ethanol, and purifying RNA with RNeasy Mini Kit For bacteria stimulation assays, fecal pellets from DSS-colitis or healthy (QIAGEN). RNA was retrotranscribed with SuperScript III First-Strand Syn- BDC2.5XNOD mice were lysed through freeze-thaw method and resus- thesis System following manufacturer’s instructions (Life Technologies). pended in RPMI-10, normalized for their protein content (Pierce BCA Protein Real-time qPCR assay was performed with SYBR Select Master Mix (Life Assay kit; Thermo Fisher), and added to total splenocytes of wt NOD mice or Technologies) using primers specific for different tight junction proteins, TCR-tg BDC2.5XNOD mice in a 96-well plate (105 cells/well) in RPMI- mucins, and cytokines (see SI Appendix, Table S1, for list of primers) on a + 10 without antibiotics. After 96 h, activation of CD4 T cells was estimated ViiA 7 Real-Time PCR System (Life Technologies). Transcript expression was + by measuring percentage of IFN-γ cells by intracellular staining. normalized against Rpl32.

Flow Cytometry. Single-cell suspensions from different organs were resus- 16S rRNA Microbiota Analysis. Total DNA was isolated from the mucosa and pended in staining buffer containing PBS, 1% FBS, and 0.09% NaN3, and luminal content of the large intestine of DSS-colitis and control BDC2.5XNOD stained with monoclonal antibodies against surface markers. For in- mice using PowerFecal DNA Isolation kit (MoBio) following the manufac- tracellular cytokine staining, single-cell suspensions isolated from different turers’ instructions with only one minor modification in lyses time, 15 min organs were stimulated for 2.5 h with 50 ng/mL phorbol 12-myristate 13- instead of 5 min, to try to retrieve all difficult-to-lyse bacteria. Microbiome μ acetate (PMA) and 1 g/mL ionomycin (both from Sigma-Aldrich) for 4 h in characterization was performed by amplification of 200 ng of extracted DNA μ the presence of 10 g/mL Brefeldin A (Sandoz). Cells were first stained for and amplification of 16S V3-5 regions using barcoded sample-specific pri- surface markers, then fixed and permeabilized using the BD Cytofix/ mers: 16S-F331 ACT CCT ACG GGA GGC AGC and 16S-R920 CCG TCA ATT Cytoperm kit, and finally stained for intracellular cytokines. The following CMT TTG AGT TT. The AccuPrime Taq DNA Polymerase (Invitrogen) and the antibodies were used: PerCP-Cy5.5 anti-mouse CD4 (RM4-5; BD Biosciences), following cycling conditions were employed: 94 °C for 2 min, 35 cycles of APC-eFluor 780 anti-mouse CD45 (30-F11; eBioscience), APC anti-mouse (94 °C for 30 s, 56 °C for 30 s, and 68 °C for 1 min), and the samples were CD25 (PC61; BioLegend), PE anti-mouse LPAM-1 (DATK32; BioLegend), FITC stored at 4 °C until usage. Amplicons were loaded on 1% agarose gel and anti-mouse Vβ4 TCR (KT4; BD Biosciences), APC-Cy7 anti-mouse CD44 (IM7; BD Biosciences), Alexa Fluor 647 anti-mouse CD62L (MEL-14; BioLegend), purified with the QiaQuick Gel Extraction Kit (Qiagen). Extracted amplicons eFluor 450 anti-mouse Foxp3 (FJK-16s; eBioscience), V450 or FITC anti-mouse were purified with AMPure XP beads (Beckman Coulter, Brea, CA) and used IFN-γ (XMG-1.2; BD Biosciences), and Alexa Fluor 647 or 488 anti-mouse for emulsion-PCR and ultradeep pyrosequencing following the 454 GS Junior ’ IL-17A (TC11-18H10; BD Biosciences). Dead cells were stained with Fixable Vi- manufacturer s instructions (Roche Diagnostics). Sequences with a high- ability Dye eFluor 506 (eBioscience) and excluded from the analysis. Flow quality score and length of >250 bp were used for the taxonomic analysis cytometry was performed using FACSCanto II (BD Biosciences), and data with Quantitative Insights Into Microbial Ecology (QIIME), version 1.6. were analyzed with FCS Express V4 software (De Novo Software). Statistical Analysis. Statistical significance of the differences between two or Fecal Transplant Experiments. Fecal microbiota transplantation (FMT) exper- more samples was calculated by unpaired two-tailed Student’s t test or iments were performed as previously described (73). The 4-wk-old ANOVA, respectively. Cumulative diabetes incidence was calculated using BDC2.5XNOD mice were depleted of their endogenous microbiota with the Kaplan–Meier estimation, whereas statistical significance was evaluated antibiotics (0.5 g/L vancomycin, 1 g/L ampicillin, 1 g/L metronidazole, and by the log-rank test. Statistical analysis of the microbiota profiling data was 1 g/L neomycin; all from Sigma-Aldrich) added to drinking water for 2 wk performed on the proportional representation of the taxa (summarized to before the FMT experiment. Three to four pellets (100 mg) of fresh stool phyla, class, order, family, and genus levels) using one-way ANOVA test with from adult donor female BDC2.5XNOD mice with/out DSS-colitis were col- Bonferroni’s correction. GraphPad Prism, version 6.0 (GraphPad Software, μ lected and pooled in 1.2 mL of sterile PBS; 200 L of the mixture was San Diego, CA), was used for statistical analysis, and values of P ≤ 0.05 were administered to recipient BDC2.5XNOD mice by oral gavage for 3 considered statistically significant. consecutive days. ACKNOWLEDGMENTS. We thank Prof. Emanuele Bosi (Ospedale San Raffaele, Diabetes Incidence. Diabetes was monitored by weekly measurements of Milan, Italy) for critical reading of the manuscript and Prof. Cecile King blood glucose levels with a GB35 Ascensia Breeze 2 glucometer (Bayer). The (The Garvan Institute of Medical Research, Sydney, NSW, Australia) for animals were considered diabetic after two consecutive blood glucose helpful suggestions. This work was supported by a research grant from the measurements ≥250 mg/dL. Juvenile Diabetes Foundation (Grant 2-SRA-2014-28-Q-R) to M.F.

1. J. A. Todd, Etiology of type 1 diabetes. Immunity 32, 457–467 (2010). 8. C. T. Brown et al., Gut microbiome metagenomics analysis suggests a functional model 2. F. Dotta et al., Coxsackie B4 virus infection of beta cells and natural killer cell insulitis in for the development of autoimmunity for type 1 diabetes. PLoS One 6, e25792 (2011). recent-onset type 1 diabetic patients. Proc. Natl. Acad. Sci. U.S.A. 104, 5115–5120 (2007). 9. T. Vatanen et al.; DIABIMMUNE Study Group, Variation in microbiome LPS immu- 3. M. C. Honeyman et al., Association between rotavirus infection and pancreatic islet nogenicity contributes to autoimmunity in humans. Cell 165, 842–853 (2016). + autoimmunity in children at risk of developing type 1 diabetes. Diabetes 49, 1319– 10. E. Badami et al., Defective differentiation of regulatory FoxP3 T cells by small- 1324 (2000). intestinal dendritic cells in patients with type 1 diabetes. Diabetes 60, 2120–2124 4. O. Vaarala et al., Cow’s milk formula feeding induces primary immunization to insulin (2011). in infants at genetic risk for type 1 diabetes. Diabetes 48, 1389–1394 (1999). 11. T. Vatanen et al., The human gut microbiome in early-onset type 1 diabetes from the 5. J. M. Norris et al., Timing of initial cereal exposure in infancy and risk of islet auto- TEDDY study. Nature 562, 589–594 (2018). immunity. JAMA 290, 1713–1720 (2003). 12. J. B. Meddings, J. Jarand, S. J. Urbanski, J. Hardin, D. G. Gall, Increased gastrointestinal 6. L. Wen et al., Innate immunity and intestinal microbiota in the development of type permeability is an early lesion in the spontaneously diabetic BB rat. Am. J. Physiol. 1 diabetes. Nature 455, 1109–1113 (2008). 276, G951–G957 (1999). 7. A. Giongo et al., Toward defining the autoimmune microbiome for type 1 diabetes. 13. S. Graham et al., Enteropathy precedes type 1 diabetes in the BB rat. Gut 53, 1437– ISME J. 5,82–91 (2011). 1444 (2004).

15148 | www.pnas.org/cgi/doi/10.1073/pnas.1814558116 Sorini et al. Downloaded by guest on October 1, 2021 14. F. Maurano et al., Small intestinal enteropathy in non-obese diabetic mice fed a diet 45. O. Vaarala, Is the origin of type 1 diabetes in the gut? Immunol. Cell Biol. 90, 271–276 containing wheat. Diabetologia 48, 931–937 (2005). (2012). 15. A. D. Mooradian, J. E. Morley, A. S. Levine, W. F. Prigge, R. L. Gebhard, Abnormal 46. A. Hänninen et al., Akkermansia muciniphila induces gut microbiota remodelling and SEE COMMENTARY intestinal permeability to sugars in diabetes mellitus. Diabetologia 29, 221–224 controls islet autoimmunity in NOD mice. Gut 67, 1445–1453 (2018). (1986). 47. H. Brauner et al., Markers of innate immune activity in patients with type 1 and type 16. R. Carratù et al., Altered intestinal permeability to mannitol in diabetes mellitus type 2 diabetes mellitus and the effect of the anti-oxidant coenzyme Q10 on inflammatory I. J. Pediatr. Gastroenterol. Nutr. 28, 264–269 (1999). activity. Clin. Exp. Immunol. 177, 478–482 (2014). 17. M. Westerholm-Ormio, O. Vaarala, P. Pihkala, J. Ilonen, E. Savilahti, Immunologic 48. D. L. Kaufman et al., Spontaneous loss of T-cell tolerance to glutamic acid de- activity in the small intestinal mucosa of pediatric patients with type 1 diabetes. Di- carboxylase in murine insulin-dependent diabetes. Nature 366,69–72 (1993). – abetes 52, 2287 2295 (2003). 49. K. Robinson, Z. Deng, Y. Hou, G. Zhang, Regulation of the intestinal barrier function 18. M. Secondulfo et al., Ultrastructural mucosal alterations and increased intestinal by host defense peptides. Front. Vet. Sci. 2, 57 (2015). – permeability in non-celiac, type I diabetic patients. Dig. Liver Dis. 36,35 45 (2004). 50. L. M. Loonen et al., REG3γ-deficient mice have altered mucus distribution and in- 19. E. Bosi et al., Increased intestinal permeability precedes clinical onset of type 1 di- creased mucosal inflammatory responses to the microbiota and enteric pathogens in abetes. Diabetologia 49, 2824–2827 (2006). the ileum. Mucosal Immunol. 7, 939–947 (2014). 20. J. Neu et al., Changes in intestinal morphology and permeability in the biobreeding 51. H. W. Koon et al., Cathelicidin signaling via the Toll-like receptor protects against rat before the onset of type 1 diabetes. J. Pediatr. Gastroenterol. Nutr. 40, 589–595 colitis in mice. Gastroenterology 141, 1852–1863.e1-3 (2011). (2005). 52. A. S. Lee et al., Gut barrier disruption by an enteric bacterial pathogen accelerates 21. O. Vaarala, M. A. Atkinson, J. Neu, The “perfect storm” for type 1 diabetes: The insulitis in NOD mice. Diabetologia 53, 741–748 (2010). complex interplay between intestinal microbiota, gut permeability, and mucosal 53. K. Berer et al., Commensal microbiota and myelin autoantigen cooperate to trigger immunity. Diabetes 57, 2555–2562 (2008). autoimmune demyelination. Nature 479, 538–541 (2011). 22. J. Sun et al., Pancreatic β-cells limit autoimmune diabetes via an immunoregulatory 54. B. D. Stadinski et al., Chromogranin A is an autoantigen in type 1 diabetes. Nat. antimicrobial peptide expressed under the influence of the gut microbiota. Immunity Immunol. 11, 225–231 (2010). 43, 304–317 (2015). 55. C. M. Gleeson, W. J. Curry, C. F. Johnston, K. D. Buchanan, Occurrence of WE-14 and 23. S. J. Turley, J. W. Lee, N. Dutton-Swain, D. Mathis, C. Benoist, Endocrine self and gut chromogranin A-derived peptides in tissues of the human and bovine gastro-entero- non-self intersect in the pancreatic lymph nodes. Proc. Natl. Acad. Sci. U.S.A. 102, – 17729–17733 (2005). pancreatic system and in human neuroendocrine neoplasia. J. Endocrinol. 151, 409 24. A. Sapone et al., Zonulin upregulation is associated with increased gut permeability in 420 (1996). subjects with type 1 diabetes and their relatives. Diabetes 55, 1443–1449 (2006). 56. W. J. Curry et al., Chromogranin A and its derived peptides in the rat and porcine 25. T. Pelaseyed et al., The mucus and mucins of the goblet cells and enterocytes provide gastro-entero-pancreatic system. Expression, localization, and characterization. Adv. the first defense line of the gastrointestinal tract and interact with the immune Exp. Med. Biol. 482, 205–213 (2000). system. Immunol. Rev. 260,8–20 (2014). 57. C. Sorini, M. Falcone, Shaping the (auto)immune response in the gut: The role of 26. M. E. Johansson, H. Sjövall, G. C. Hansson, The gastrointestinal mucus system in health intestinal immune regulation in the prevention of type 1 diabetes. Am. J. Clin. Exp. and disease. Nat. Rev. Gastroenterol. Hepatol. 10, 352–361 (2013). Immunol. 2, 156–171 (2013). 27. P. M. Smith et al., The microbial metabolites, short-chain fatty acids, regulate colonic 58. M. Knip, H. Siljander, The role of the intestinal microbiota in type 1 diabetes mellitus. – Treg cell homeostasis. Science 341, 569–573 (2013). Nat. Rev. Endocrinol. 12, 154 167 (2016). INFLAMMATION 28. N. Arpaia et al., Metabolites produced by commensal bacteria promote peripheral 59. J. D. Taurog et al., The germfree state prevents development of gut and joint in- IMMUNOLOGY AND regulatory T-cell generation. Nature 504, 451–455 (2013). flammatory disease in HLA-B27 transgenic rats. J. Exp. Med. 180, 2359–2364 (1994). 29. R. L. Gallo, L. V. Hooper, Epithelial antimicrobial defence of the skin and intestine. 60. C. King, N. Sarvetnick, The incidence of type-1 diabetes in NOD mice is modulated by Nat. Rev. Immunol. 12, 503–516 (2012). restricted flora not germ-free conditions. PLoS One 6, e17049 (2011). 30. P. Chairatana, E. M. Nolan, Defensins, lectins, mucins, and secretory immunoglobulin 61. M. W. Sadelain, H. Y. Qin, J. Lauzon, B. Singh, Prevention of type I diabetes in NOD A: Microbe-binding biomolecules that contribute to mucosal immunity in the human mice by adjuvant immunotherapy. Diabetes 39, 583–589 (1990). gut. Crit. Rev. Biochem. Mol. Biol. 52,45–56 (2017). 62. T. C. Martins, A. P. Aguas, Changes in B and T lymphocytes associated with 31. M. Shan et al., Mucus enhances gut homeostasis and oral tolerance by delivering mycobacteria-induced protection of NOD mice from diabetes. J. Autoimmun. 9, 501– immunoregulatory signals. Science 342, 447–453 (2013). 507 (1996). 32. Y. H. Sheng et al., MUC1 and MUC13 differentially regulate epithelial inflammation in 63. D. Elias, D. Markovits, T. Reshef, R. van der Zee, I. R. Cohen, Induction and therapy of – response to inflammatory and infectious stimuli. Mucosal Immunol. 6, 557 568 autoimmune diabetes in the non-obese diabetic (NOD/Lt) mouse by a 65-kDa heat (2013). shock protein. Proc. Natl. Acad. Sci. U.S.A. 87, 1576–1580 (1990). 33. P. Courtois, C. Jurysta, A. Sener, F. W. Scott, W. J. Malaisse, Quantitative and quali- 64. P. Saï, A. S. Rivereau, Prevention of diabetes in the nonobese diabetic mouse by oral – tative alterations of intestinal mucins in BioBreeding rats. Int. J. Mol. Med. 15, 105 immunological treatments. Comparative efficiency of human insulin and two bacte- 108 (2005). rial antigens, lipopolysacharide from Escherichia coli and glycoprotein extract from 34. I. Jaakkola, S. Jalkanen, A. Hänninen, Diabetogenic T cells are primed both in pan- Klebsiella pneumoniae. Diabetes Metab. 22, 341–348 (1996). creatic and gut-associated lymph nodes in NOD mice. Eur. J. Immunol. 33, 3255–3264 65. R. Horai et al., Microbiota-dependent activation of an autoreactive T cell receptor (2003). provokes autoimmunity in an immunologically privileged site. Immunity 43, 343–353 35. J. D. Katz, B. Wang, K. Haskins, C. Benoist, D. Mathis, Following a diabetogenic T cell (2015). from genesis through pathogenesis. Cell 74, 1089–1100 (1993). 66. T. C. Fung et al., Lymphoid-tissue-resident commensal bacteria promote members of 36. K. Haskins, M. Portas, B. Bradley, D. Wegmann, K. Lafferty, T-lymphocyte clone spe- the IL-10 cytokine family to establish mutualism. Immunity 44, 634–646 (2016). cific for pancreatic islet antigen. Diabetes 37, 1444–1448 (1988). 67. D. K. Cole et al., Hotspot autoimmune T cell receptor binding underlies pathogen and 37. R. Mueller, L. M. Bradley, T. Krahl, N. Sarvetnick, Mechanism underlying counter- insulin peptide cross-reactivity. J. Clin. Invest. 126, 3626 (2016). regulation of autoimmune diabetes by IL-4. Immunity 7, 411–418 (1997). 68. S. Wirtz, C. Neufert, B. Weigmann, M. F. Neurath, Chemically induced mouse models 38. M. S. Horwitz et al., Diabetes induced by Coxsackie virus: Initiation by bystander – damage and not molecular mimicry. Nat. Med. 4, 781–785 (1998). of intestinal inflammation. Nat. Protoc. 2, 541 546 (2007). 39. M. E. Johansson et al., Bacteria penetrate the inner mucus layer before inflammation 69. R. L. Brown, R. P. Sequeira, T. B. Clarke, The microbiota protects against respiratory in the dextran sulfate colitis model. PLoS One 5, e12238 (2010). infection via GM-CSF signaling. Nat. Commun. 8, 1512 (2017). 40. C. Hernández-Chirlaque et al., Germ-free and antibiotic-treated mice are highly sus- 70. K. Atarashi et al., Induction of colonic regulatory T cells by indigenous Clostridium – ceptible to epithelial injury in DSS colitis. J. Crohns Colitis 10, 1324–1335 (2016). species. Science 331, 337 341 (2011). 41. N. Tai et al., Microbial antigen mimics activate diabetogenic CD8 T cells in NOD mice. 71. B. Weigmann et al., Isolation and subsequent analysis of murine lamina propria J. Exp. Med. 213, 2129–2146 (2016). mononuclear cells from colonic tissue. Nat. Protoc. 2, 2307–2311 (2007). 42. F. R. Costa et al., Gut microbiota translocation to the pancreatic lymph nodes triggers 72. L. Lefrancois, N. Lycke, Isolation of mouse small intestinal intraepithelial lymphocytes, NOD2 activation and contributes to T1D onset. J. Exp. Med. 213, 1223–1239 (2016). Peyer’s patch, and lamina propria cells. Curr. Protoc. Immunol. Chapter 3, Unit 3.19 43. X. Li, M. A. Atkinson, The role for gut permeability in the pathogenesis of type 1 di- (2001). abetes—a solid or leaky concept? Pediatr. Diabetes 16, 485–492 (2015). 73. S. Ghosh et al., Colonic microbiota alters host susceptibility to infectious colitis by 44. O. Vaarala, Leaking gut in type 1 diabetes. Curr. Opin. Gastroenterol. 24, 701–706 modulating inflammation, redox status, and ion transporter gene expression. Am. J. (2008). Physiol. Gastrointest. Liver Physiol. 301, G39–G49 (2011).

Sorini et al. PNAS | July 23, 2019 | vol. 116 | no. 30 | 15149 Downloaded by guest on October 1, 2021